meaning of majority rule (was Re: N factorial combinations (was

[Mike Ossipoff]

Steve Eppley writes:

[Mike says: I can't find the original letter from Demorep, so
I must reply to this reply]

> 
> Demorep wrote:
> -snip-
> >As Mr. Arrow informs us, choosing the *best* method involves
> >various tradeoffs in the strategies regarding the various N1....
> >Nmost values.  I will maintain that majority rule must be an
> >absolute requirement in any method.

Demorep is still telling us about what Mr. Arrow informs us, but,
as usual, he neglects to cite specific statements by Arrow, but,
instead, seems to be pretty much making it up as he goes along.
Maybe Arrow is whatever Demorep wants him to be.

Demorep's judgement of majority rule, based on absolute y/n
votes is quite irrelevant to a multicandidate election, where
the choice involves _relative_ preference, not absolute. Absolute
preferences are simply meaningless in a multicandidate election.
A choice between several candidates, even if one of them is
"no winner; hold a new election"(for candidates) or
"status quo" (for initiative alternatives) is a _relative_
choice. Demorep still doesn't understand that.

And in trying to ensure majority rule for the meaningless absolute
y/n votes, Demorep's proposal sacrifices it in regards to relative
preferences. But why am I wasting my time answering that crap.

> 
> Different people define majority rule in different ways, and in many
> cases don't really define the term at all.  
> 
> The GMC and NML criteria are two definitions which Condorcet
> supports.  MPV and Coombs advocates might argue that those methods 
> do support a narrow definition of majority rule, since they do pay
> attention to majority in 1 of the N(N-1)/2 candidate pairings. 

But MPV is much worse than Coombs, and that's why it would be good
if EM's only MPV advocate switched to Coombs.

Coombs is better than MPV because of its LO2E-2 compliance, and
because of its much less demanding strategy, when there's
a 1-dimensional political spectrum.

> 
> ---Steve     (Steve Eppley    seppley at alumni.caltech.edu)
> 
> .-
> 


--